Note
Click here to download the full example code
Layer NormalizationΒΆ
Out:
layer-norm-backward:
N Triton Torch Apex
0 1024.0 307.200008 99.902435 307.200008
1 1536.0 351.085717 135.032961 341.333333
2 2048.0 420.102553 162.754967 327.679984
3 2560.0 458.507457 182.857144 330.322572
4 3072.0 515.580429 191.501303 319.168834
5 3584.0 547.872604 207.768111 311.652167
6 4096.0 568.231237 221.905193 301.546004
7 4608.0 504.986315 232.336141 287.999990
8 5120.0 531.948056 242.366855 285.104413
9 5632.0 538.517949 243.545956 290.683877
10 6144.0 546.133354 250.349744 288.000001
11 6656.0 536.053693 256.000009 286.279570
12 7168.0 510.480705 252.988236 277.024148
13 7680.0 482.513091 267.130429 284.444450
14 8192.0 463.698115 269.326017 282.482757
15 8704.0 417.791980 264.425310 282.673891
16 9216.0 431.157889 274.762727 291.031570
17 9728.0 439.683593 281.630872 290.027323
18 10240.0 446.025405 286.100109 289.811322
19 10752.0 425.120247 246.935876 289.941565
20 11264.0 425.056596 243.765566 283.371073
21 11776.0 423.089806 249.888595 289.129414
22 12288.0 421.302872 254.673582 295.207195
23 12800.0 414.574901 254.515329 290.909089
24 13312.0 414.381327 253.561895 289.129403
25 13824.0 406.090579 257.790206 293.088338
26 14336.0 394.116833 256.381525 290.349381
27 14848.0 385.245405 255.999999 287.844912
28 15360.0 376.932517 262.751252 291.184839
29 15872.0 370.913333 262.166551 291.118085
import torch
import triton
import triton.language as tl
try:
# This is https://github.com/NVIDIA/apex, NOT the apex on PyPi, so it
# should not be added to extras_require in setup.py.
import apex
HAS_APEX = True
except ModuleNotFoundError:
HAS_APEX = False
# Forward Pass
@triton.jit
def _layer_norm_fwd_fused(X, Y, W, B, M, V, stride, N, eps,
BLOCK_SIZE: tl.constexpr):
# position of elements processed by this program
row = tl.program_id(0)
cols = tl.arange(0, BLOCK_SIZE)
mask = cols < N
# offset data pointers to start at the row of interest
X += row * stride
Y += row * stride
# load data and cast to float32
x = tl.load(X + cols, mask=mask, other=0).to(tl.float32)
# compute mean
mean = tl.sum(x, axis=0) / N
# compute std
xmean = tl.where(mask, x - mean, 0.)
var = tl.sum(xmean * xmean, axis=0) / N
rstd = 1 / tl.sqrt(var + eps)
xhat = xmean * rstd
# write-back mean/rstd
tl.store(M + row, mean)
tl.store(V + row, rstd)
# multiply by weight and add bias
w = tl.load(W + cols, mask=mask)
b = tl.load(B + cols, mask=mask)
y = xhat * w + b
# write-back
tl.store(Y + cols, y, mask=mask)
# Backward pass (DX + partial DW + partial DB)
@triton.jit
def _layer_norm_bwd_dx_fused(DX, DY, DW, DB, X, W, B, M, V, Lock, stride, N, eps,
GROUP_SIZE_M: tl.constexpr, BLOCK_SIZE_N: tl.constexpr):
# position of elements processed by this program
row = tl.program_id(0)
cols = tl.arange(0, BLOCK_SIZE_N)
mask = cols < N
# offset data pointers to start at the row of interest
X += row * stride
DY += row * stride
DX += row * stride
# offset locks and weight/bias gradient pointer
# each kernel instance accumulates partial sums for
# DW and DB into one of GROUP_SIZE_M independent buffers
# these buffers stay in the L2, which allow this kernel
# to be fast
lock_id = row % GROUP_SIZE_M
Lock += lock_id
Count = Lock + GROUP_SIZE_M
DW = DW + lock_id * N + cols
DB = DB + lock_id * N + cols
# load data to SRAM
x = tl.load(X + cols, mask=mask, other=0).to(tl.float32)
dy = tl.load(DY + cols, mask=mask, other=0).to(tl.float32)
w = tl.load(W + cols, mask=mask).to(tl.float32)
mean = tl.load(M + row)
rstd = tl.load(V + row)
# compute dx
xhat = (x - mean) * rstd
wdy = w * dy
xhat = tl.where(mask, xhat, 0.)
wdy = tl.where(mask, wdy, 0.)
mean1 = tl.sum(xhat * wdy, axis=0) / N
mean2 = tl.sum(wdy, axis=0) / N
dx = (wdy - (xhat * mean1 + mean2)) * rstd
# write-back dx
tl.store(DX + cols, dx, mask=mask)
# accumulate partial sums for dw/db
partial_dw = (dy * xhat).to(w.dtype)
partial_db = (dy).to(w.dtype)
while tl.atomic_cas(Lock, 0, 1) == 1:
pass
count = tl.load(Count)
# first store doesn't accumulate
if count == 0:
tl.atomic_xchg(Count, 1)
else:
partial_dw += tl.load(DW, mask=mask)
partial_db += tl.load(DB, mask=mask)
tl.store(DW, partial_dw, mask=mask)
tl.store(DB, partial_db, mask=mask)
# release lock
tl.atomic_xchg(Lock, 0)
# Backward pass (total DW + total DB)
@triton.jit
def _layer_norm_bwd_dwdb(DW, DB, FINAL_DW, FINAL_DB, M, N,
BLOCK_SIZE_M: tl.constexpr, BLOCK_SIZE_N: tl.constexpr):
pid = tl.program_id(0)
cols = pid * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N)
dw = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
db = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
for i in range(0, M, BLOCK_SIZE_M):
rows = i + tl.arange(0, BLOCK_SIZE_M)
mask = (rows[:, None] < M) & (cols[None, :] < N)
offs = rows[:, None] * N + cols[None, :]
dw += tl.load(DW + offs, mask=mask, other=0.)
db += tl.load(DB + offs, mask=mask, other=0.)
sum_dw = tl.sum(dw, axis=0)
sum_db = tl.sum(db, axis=0)
tl.store(FINAL_DW + cols, sum_dw, mask=cols < N)
tl.store(FINAL_DB + cols, sum_db, mask=cols < N)
class LayerNorm(torch.autograd.Function):
@staticmethod
def forward(ctx, x, normalized_shape, weight, bias, eps):
# allocate output
y = torch.empty_like(x)
# reshape input data into 2D tensor
x_arg = x.reshape(-1, x.shape[-1])
M, N = x_arg.shape
mean = torch.empty((M, ), dtype=torch.float32, device='cuda')
rstd = torch.empty((M, ), dtype=torch.float32, device='cuda')
# Less than 64KB per feature: enqueue fused kernel
MAX_FUSED_SIZE = 65536 // x.element_size()
BLOCK_SIZE = min(MAX_FUSED_SIZE, triton.next_power_of_2(N))
if N > BLOCK_SIZE:
raise RuntimeError("This layer norm doesn't support feature dim >= 64KB.")
# heuristics for number of warps
num_warps = min(max(BLOCK_SIZE // 256, 1), 8)
# enqueue kernel
_layer_norm_fwd_fused[(M,)](x_arg, y, weight, bias, mean, rstd,
x_arg.stride(0), N, eps,
BLOCK_SIZE=BLOCK_SIZE, num_warps=num_warps)
ctx.save_for_backward(x, weight, bias, mean, rstd)
ctx.BLOCK_SIZE = BLOCK_SIZE
ctx.num_warps = num_warps
ctx.eps = eps
return y
@staticmethod
def backward(ctx, dy):
x, w, b, m, v = ctx.saved_tensors
# heuristics for amount of parallel reduction stream for DG/DB
N = w.shape[0]
GROUP_SIZE_M = 64
if N <= 8192: GROUP_SIZE_M = 96
if N <= 4096: GROUP_SIZE_M = 128
if N <= 1024: GROUP_SIZE_M = 256
# allocate output
locks = torch.zeros(2 * GROUP_SIZE_M, dtype=torch.int32, device='cuda')
_dw = torch.empty((GROUP_SIZE_M, w.shape[0]), dtype=x.dtype, device=w.device)
_db = torch.empty((GROUP_SIZE_M, w.shape[0]), dtype=x.dtype, device=w.device)
dw = torch.empty((w.shape[0],), dtype=w.dtype, device=w.device)
db = torch.empty((w.shape[0],), dtype=w.dtype, device=w.device)
dx = torch.empty_like(dy)
# enqueue kernel using forward pass heuristics
# also compute partial sums for DW and DB
x_arg = x.reshape(-1, x.shape[-1])
M, N = x_arg.shape
_layer_norm_bwd_dx_fused[(M,)](dx, dy, _dw, _db, x, w, b, m, v, locks,
x_arg.stride(0), N, ctx.eps,
BLOCK_SIZE_N=ctx.BLOCK_SIZE,
GROUP_SIZE_M=GROUP_SIZE_M,
num_warps=ctx.num_warps)
grid = lambda meta: [triton.cdiv(N, meta['BLOCK_SIZE_N'])]
# accumulate partial sums in separate kernel
_layer_norm_bwd_dwdb[grid](_dw, _db, dw, db, GROUP_SIZE_M, N,
BLOCK_SIZE_M=32,
BLOCK_SIZE_N=128)
return dx, None, dw, db, None
layer_norm = LayerNorm.apply
def test_layer_norm(M, N, dtype, eps=1e-5, device='cuda'):
# create data
x_shape = (M, N)
w_shape = (x_shape[-1], )
weight = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
bias = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
x = -2.3 + 0.5 * torch.randn(x_shape, dtype=dtype, device='cuda')
dy = .1 * torch.randn_like(x)
x.requires_grad_(True)
# forward pass
y_tri = layer_norm(x, w_shape, weight, bias, eps)
y_ref = torch.nn.functional.layer_norm(x, w_shape, weight, bias, eps).to(dtype)
# backward pass (triton)
y_tri.backward(dy, retain_graph=True)
dx_tri, dw_tri, db_tri = [_.grad.clone() for _ in [x, weight, bias]]
x.grad, weight.grad, bias.grad = None, None, None
# backward pass (torch)
y_ref.backward(dy, retain_graph=True)
dx_ref, dw_ref, db_ref = [_.grad.clone() for _ in [x, weight, bias]]
# compare
triton.testing.assert_almost_equal(y_tri, y_ref)
triton.testing.assert_almost_equal(dx_tri, dx_ref)
triton.testing.assert_almost_equal(db_tri, db_ref, decimal=1)
triton.testing.assert_almost_equal(dw_tri, dw_ref, decimal=1)
@triton.testing.perf_report(
triton.testing.Benchmark(
x_names=['N'],
x_vals=[512 * i for i in range(2, 32)],
line_arg='provider',
line_vals=['triton', 'torch'] + (['apex'] if HAS_APEX else []),
line_names=['Triton', 'Torch'] + (['Apex'] if HAS_APEX else []),
styles=[('blue', '-'), ('green', '-'), ('orange', '-')],
ylabel='GB/s',
plot_name='layer-norm-backward',
args={'M': 4096, 'dtype': torch.float16, 'mode': 'backward'}
)
)
def bench_layer_norm(M, N, dtype, provider, mode='backward', eps=1e-5, device='cuda'):
# create data
x_shape = (M, N)
w_shape = (x_shape[-1], )
weight = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
bias = torch.rand(w_shape, dtype=dtype, device='cuda', requires_grad=True)
x = -2.3 + 0.5 * torch.randn(x_shape, dtype=dtype, device='cuda')
dy = .1 * torch.randn_like(x)
x.requires_grad_(True)
# utility functions
if provider == 'triton':
y_fwd = lambda: layer_norm(x, w_shape, weight, bias, eps)
if provider == 'torch':
y_fwd = lambda: torch.nn.functional.layer_norm(x, w_shape, weight, bias, eps)
if provider == 'apex':
apex_layer_norm = apex.normalization.FusedLayerNorm(w_shape).to(x.device).to(x.dtype)
y_fwd = lambda: apex_layer_norm(x)
# forward pass
if mode == 'forward':
gbps = lambda ms: 2 * x.numel() * x.element_size() / ms * 1e-6
ms, min_ms, max_ms = triton.testing.do_bench(y_fwd, rep=500)
# backward pass
if mode == 'backward':
gbps = lambda ms: 3 * x.numel() * x.element_size() / ms * 1e-6
y = y_fwd()
ms, min_ms, max_ms = triton.testing.do_bench(lambda: y.backward(dy, retain_graph=True),
grad_to_none=[x], rep=500)
return gbps(ms), gbps(max_ms), gbps(min_ms)
bench_layer_norm.run(save_path='.', print_data=True)
Total running time of the script: ( 2 minutes 11.560 seconds)